Skip to documentation content
Dataset APIDocs

Dataset API introduction

Publish and consume versioned datasets without coupling integrations to builder internals.

In this guide
  1. 1Understand the resource model
  2. 2Trust immutable releases
  3. 3Choose query or download
In this guide
  1. 1Understand the resource model
  2. 2Trust immutable releases
  3. 3Choose query or download

One stable contract over an evolving builder

Dataset API separates customer integrations from the Python dataset_builder implementation. Consumers use a versioned HTTPS contract while collectors, normalizers, and validation logic can evolve behind the adapter.

The public workflow is intentionally small: discover a dataset, inspect or create a build, select an immutable release, then query bounded rows or download the release artifact.

Resource map

ResourcePurposeMutability
DatasetCatalog definition, fields, identity, and latest releaseDefinition may gain compatible fields
BuildAsynchronous builder jobStatus changes until terminal
ReleaseQuality-checked published snapshotImmutable
Query responseBounded projection from one releaseRead-only
DownloadFull release artifactRead-only

The release is the unit of trust

A release identifies its dataset and schema versions, quality status, counts, optional parent release, artifacts, and immutable flag. Persist the release ID with every downstream import so results remain reproducible.

First managed catalog

The preview catalog contains Finnish vehicle market listings. The narrow starting point makes schema evolution, source coverage, carried-forward rows, and release lineage concrete before additional catalogs are introduced.

Was this page helpful?